yes Visual recognition and coevolutionary history drive responses of amphibians to an invasive predator [1/1, 1 ed.]

Citation preview

Behavioral Ecology

The official journal of the

ISBE

International Society for Behavioral Ecology

Behavioral Ecology (2021), 32(6), 1352–1362. https://doi.org/10.1093/beheco/arab101

Original Article

Visual recognition and coevolutionary history drive responses of amphibians to an invasive predator Andrea Melotto,a,b, Gentile Francesco Ficetola,a,c, Elisa Alari,a Samuele Romagnoli,a and Raoul Manentia, aDepartment of Environmental Science and Policy, Università degli Studi di Milano, Milan 20133, Italy, bCentre of Excellence for Invasion Biology, Department of Botany and Zoology, Stellenbosch University, Stellenbosch, 7600, South Africa, and cLaboratoire D’Ecologie Alpine (LECA), CNRS, Université de Grenoble Alpes, Grenoble 38000, France Received 25 November 2020; revised 31 May 2021; editorial decision 29 July 2021; accepted 23 August 2021; Advance Access publication 26 September 2021.

During biotic invasions, native prey are abruptly exposed to novel predators and are faced with unprecedented predatory pressures. Under these circumstances, the lack of common evolutionary history may hamper predator recognition by native prey, undermining the expression of effective antipredator responses. Nonetheless, mechanisms allowing prey to overcome evolutionary naïveté exist. For instance, in naïve prey, history of coevolution with similar native predators or detection of general traits characterizing predators can favor the recognition of stimuli released by invasive predators. However, few studies have assessed how these mechanisms shape prey response at the community level. Here, we evaluated behavioral responses in naïve larvae of 13 amphibian species to chemical and visual cues associated with an invasive predator, the American red swamp crayfish (Procambarus clarkii). Moreover, we investigated how variation among species responses was related to their coexistence with similar native crayfish predators. Amphibian larvae altered their behavior in presence of visual stimuli of the alien crayfish, while chemical cues elicited feeble and contrasting behavioral shifts. Activity reduction was the most common and stronger response, whereas some species exhibited more heterogeneous strategies also involving distancing and rapid escape response. Interestingly, species sharing coevolutionary history with the native crayfish were able to finely tune their response to the invasive one, performing bursts to escape. These results suggest native prey can respond to invasive predators through recognition of generic risk cues (e.g., approaching large shapes), still the capability of modulating antipredator strategies may also depend on their coevolutionary history with similar native predators. Key words:  amphibian community, antipredator behavior, history of coexistence, invasive species, predator recognition, prey naïveté.

INTRODUCTION Biotic invasions are increasingly shaping ecosystems at the global scale and constitute one of the major drivers of biodiversity loss (Mooney and Cleland 2001; Clavero and Garcia-Berthou 2005; Bellard et al. 2016). Invasive predators have severe impacts on invaded ecosystems, often leading to sharp declines and local extinction of native prey populations worldwide (Rodda et al. 1997; Kats and Ferrer 2003; Salo et al. 2007; Cruz et al. 2008; Doherty et al. 2016), as they expose native species to novel and abrupt predation pressures (Cox and Lima 2006; Sih et al. 2010; Carthey and Banks  

Address correspondence to A. Melotto. E-mail: [email protected].

2014). Under these circumstances, behavioral responses can be extremely important, as they can constitute a first line of defense for native species towards invasive ones (Holway and Suarez 1999; Weis and Sol 2016). Correct risk assessment is crucial for prey as it is required to foster effective antipredator responses and finely tune their expression according to the perceived risk (Lima and Dill 1990; Lima and Bednekoff 1999; Ferrari and Chivers 2011). Predator recognition can be mediated by a wide variety of stimuli (Lima and Dill 1990), which depend on the ecological context wherein prey species have evolved, and is favored by the presence of a history of coevolution between predator and prey (Downes and Shine 1998; Sih et  al. 2010). Thus, when a non-native predator invades an ecosystem, crucial questions arise on prey capability

© The Author(s) 2021. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: [email protected]

Melotto et al. • Visual cues and coevolution drive antipredator responses

to withstand the novel threat. How naïve prey respond to the new threat and how responses vary across native prey community? Which mechanisms can favor novel predator recognition? Native prey can fail to perceive invasive predators as a potential threat or fail to associate cues they release to predation risk, and this generally hampers the expression of adequate antipredator responses (Salo et  al. 2007; Gomez-Mestre and Díaz-Paniagua 2011). Failed predator recognition in native prey is often attributed to the lack of common evolutionary history with the invasive species (Cox and Lima 2006; Sih et  al. 2010). This lack of responsiveness due to absence of coevolutionary history is known as evolutionary naïveté (Carthey and Banks 2014; Carthey and Blumstein 2018). However, mechanisms allowing to overcome evolutionary naïveté in prey exist (Cox and Lima 2006; Carthey and Banks 2014) and in some cases native prey can recognize novel predators. On the one hand, when the invasive predator is phylogenetically close or shares similar traits with a native predator, prey can recognize predator archetype and broaden their antipredator response to the novel species (predator generalization hypothesis) (Griffin et  al. 2001; Ferrari et  al. 2007; Davis et  al. 2012). In other cases, a novel species can be “labeled” as a predator by naïve prey when it shares traits that are commonly associated with predator species (e.g., large size; stealthy approaching) (Carthey and Blumstein 2018), inducing a generic antipredator response in prey (generic response hypothesis) (Mathis and Vincent 2000; Rehage et  al. 2009; Wilson et al. 2018). Moreover, while generic responses are commonly triggered by visual cues (Mathis and Vincent 2000), predator generalization often involves chemical stimuli, as phylogenetically close predators tend to produce similar kairomones (i.e., predator chemical cues) that can be recognized by native prey (Ferrari et  al. 2007; Davis et  al. 2012). Finally, it is worth remarking that other mechanisms, such as learning through the association of novel stimuli to familiar risk cues (Gonzalo et  al. 2007; Nunes et  al. 2013) or neophobia (Chivers et  al. 2014), may also be involved in shaping or modulating prey responses to novel predators. Freshwater systems are closely connected habitats that are highly vulnerable to disturbances at network scale, such as damming and fragmentation, and are strongly exposed to invasive species (Leprieur et  al. 2008; Strayer 2010). In these habitats, native prey naïveté to introduced predators can thus be particularly frequent (Cox and Lima 2006; Rehage et al. 2009). In aquatic environments, visual stimuli and chemical communication are major cues used by prey for risk assessment (Chivers et  al. 2001; Wisenden 2003; Ferrari et  al. 2010c; Hettyey et  al. 2012). Visual cues primarily allow to locate predators and are involved in rapid predator avoidance (Hettyey et  al. 2012), but they can also contribute to refine risk assessment and discriminate between predators actually constituting a threat and nonthreatening predators (e.g., by assessing predator size) (Chivers et  al. 2001). Nonetheless, freshwater environments frequently have poor visibility (e.g., turbid or densely vegetated water), thus visual recognition is often useful only at short distances and cannot prevent predator encounters (Abrahams and Kattenfeld 1997; Ferrari et  al. 2010b). Conversely, chemical cues can be perceived before encountering the predator and can elicit antipredator responses aimed at preventing exposure to predators (Kats and Dill 1998). Chemical stimuli can also provide information on predator diet and density (Schoeppner and Relyea 2005, 2008), allowing prey to finely tune antipredator response on the basis of actual predation risk (Benard 2006).

1353

Antipredator responses against novel predators can have key consequences on the dynamics of invaded communities. Native species recognizing invasive predators as a threat can exhibit more effective antipredator responses, and this could increase their ability to withstand the impact of invaders. However, understanding inter-specific variation of antipredator responses can be challenging, because it requires the comparison of a large number of species, potential stimuli, and potential responses, and thus very few studies have so far assessed the antipredator responses to invasive predators at the community level (but see Rebelo and Cruz 2005; Nunes et al. 2013; Nunes et al. 2014a). Here we investigated the capability to recognize a non-native predator and express behavioral responses across the 13 species composing the amphibian communities of freshwaters in Northern Italy. During behavioral tests, we monitored variation of activity and space use in naïve amphibian larvae exposed to a combination of visual and chemical stimuli from an invasive predator, the American red swamp crayfish Procambarus clarkii (hereafter American crayfish), which is a major threat to freshwater biodiversity (Nentwig et  al. 2018). In so doing, we aimed to assess (i) how the response to the alien predator varies among species; (ii) what is the relative role of predator-released stimuli (i.e., visual and chemical cues) in mediating risk assessment and antipredator behavior in native amphibian prey; (iii) if interspecific variation in antipredator responses can be explained by the generalization hypothesis, or by the generic response hypothesis. The generalization hypothesis predicts better antipredator responses in amphibians that co-evolved with a similar native predator (i.e., the European white-clawed crayfish, Austropotamobius pallipes; hereafter European crayfish), while the generic response hypothesis predicts comparable responses across species.

METHODS Study area and animal sampling We considered 13 amphibian species, which represent the most common pond-breeding amphibian species in Northern Italy. The study species included five caudates and eight anurans: fire salamander (Salamandra salamandra) northern spectacled salamander (Salamandrina perspicillata), smooth newt (Lissotriton vulgaris), Italian crested newt (Triturus carnifex), alpine newt (Ichthyosaura alpestris), Italian agile frog (Rana latastei), agile frog (Rana dalmatina), Italian stream frog (Rana italica), European common frog (Rana temporaria), green frog (Pelophylax kl. esculentus), Italian tree frog (Hyla intermedia), European common toad (Bufo bufo), and European green toad (Bufo viridis complex). All the study species were collected in the Po River Valley or in the Northern Apennines (administrative regions: Lombardia, Liguria, and Emilia Romagna; see Supplementary Figure S1). This area hosts a rich hydrographic network where broadleaved forests are intermingled with urban and agricultural areas. In these regions, the native European crayfish, Austropotamobius pallipes, which is an amphibian predator generally living in small streams, was historically common (Manenti et  al. 2014). Nonetheless, the European crayfish has undergone a rapid decline in the last century, due to habitat modification, fishing, and spread of pathogens, and is now extinct in most of its historical range (Holdich et al. 2009; Bonelli et al. 2017; Manenti et  al. 2019). To test if the coevolutionary history with the native crayfish allows amphibians to respond towards invasive crayfish, we selected amphibian populations breeding in sites within hydrographic basins that hosted the European crayfish in the past.

Behavioral Ecology

1354

Between spring and summer 2018, we collected 12 larvae from two populations for each of the 13 amphibian species (total: 26 populations, 312 individuals). Amphibian larvae were at intermediate developmental stages (for anurans, Gosner’s stage 28–33 (Gosner 1960); for caudates, stages 51b-52b according to (Bernabò and Brunelli 2019)) and were all collected from populations where the European crayfish is extinct since a decade (amphibians sharing a coevolutionary history with the native crayfish) or naturally absent (amphibians without coevolutionary history). All amphibian larvae came from populations uninvaded by the alien crayfish. This allowed to exclude potential effects of individual experience towards any crayfish predator, or the possibility of a recent evolutionary response to the invasive crayfish. Procambarus clarkii is native of North America but is currently widespread in Northern Italy, even if its distribution is patchy (Lo Parrino et al. 2020). This invasive crayfish has a broad niche and is able to exploit both rivers and lentic environments (Souty-Grosset et  al. 2006). The overall morphology and the predatory behavior is similar between the invasive and the European crayfish, even though the invasive one shows a more opportunistic diet and has a better ability to capture prey (Gherardi et  al. 2001; Renai and Gherardi 2004). As a consequence, several amphibian populations invaded by the American crayfish underwent recent declines (Cruz et al. 2008; Falaschi et al. 2021), and in some cases, the selective pressure posed by this crayfish was strong enough to even trigger rapid adaptation in invaded populations of some species (Melotto et  al. 2020). American crayfish individuals used in this study (n = 40) were collected from a dense population in Lombardy (approx. 45.729°N, 9.237°E).

Housing and experimental protocol After collection, larvae were housed in laboratory within 49 x 35  cm plastic tanks containing 15  L of decanted tap water. Each tank hosted 12 larvae from the same population, which were individually housed in perforated plastic cups (Ø  =  8  cm). Larvae were kept under constant oxygenation, and were exposed to room temperature and daily photoperiod. During their housing period, larvae were fed every second day with rabbit pellet (anuran tadpoles), Chironomus spp. larvae (Salamandra larvae), or Daphnia spp. (Salamandrina and newt larvae). After collection, P.  clarkii individuals (cephalothorax length: mean ± SE = 46.92 ± 0.75 mm) were singularly hosted in plastic tanks (20 x 14 cm, 5 L of decanted tap water), in the same conditions as amphibian larvae and fed with commercial fish food every second day. All larvae were housed in the lab for a minimum of three days before behavioral tests (mean ± SE: 4.7  ± 1 d). After a two-day starvation period, we performed one experimental session for each amphibian population (i.e., 26 experimental sessions in total). During experimental sessions, each amphibian larva was exposed to the non-lethal presence of the American crayfish with four combinations of cues deriving from the predator (Figure 1): visual and chemical cues (V+C+); visual cues only (V+C–); chemical cues only (V–C+); no risk cues (V–C–). Experiments were conducted in 51  × 18  cm plastic tanks, filled with 8  L of decanted tap water. Experimental tanks were divided into two compartments by a transparent plastic barrier. This barrier was impermeable to water and any unintended chemical cue exchange between the two compartments was avoided. One compartment hosted amphibian larvae (18 x 18 cm, hereafter “prey compartment”), while the second one hosted the American crayfish (32  × 18  cm, hereafter “predator compartment”). In all predator compartments,

an opaque plastic pot (9  × 9  × 14  cm) was present. Pots hosted the crayfish in treatments without visual cues, while in visual-cue treatments the crayfish was free ranging in its compartment. In treatments with exposure to P.  clarkii chemical cues, 0.5  cm diameter holes, performed both on the barrier separating larvae from the invasive crayfish (n  =  15) and the pot (n  =  12 per each of the two lateral sides), allowed chemical cue exchange between compartments. Behavioral tests were conducted between 7 AM. and 7 PM.; all individuals from the same population were tested on the same day, while different populations were tested separately. Before experiments started, each larva was inserted in the prey compartment and let acclimatize for 3  min. After acclimatization, we inserted a crayfish in the predator compartment (in the pot for V–C+ and V–C– treatments; out of the pot for V+C+, V+C– treatments). Behavioral tests lasted 7 min and larva activity was video-recorded by placing a Nikon d5300 camera (18  mm lens) perpendicularly above the prey compartment. For each individual, we performed eight behavioral tests (four treatments, each replicated twice). Tests were conducted in a randomized order to minimize the potential bias of exposure sequence (Altmann 1974; Ferrari et  al. 2010a; Melotto et al. 2019). During each experimental day, 12 crayfish individuals were selected and assigned to behavioral trials following a randomized protocol, so that each crayfish was used twice for the same condition. We left at least 15  min recovering time for each animal between consecutive tests. Each tank and pot were assigned to a treatment and then used for that specific treatment only. Tanks and pots were washed multiple times between subsequent trials to minimize traces of cues from preceding tests. In total, we performed 2496 behavioral tests (12 individuals × 26 populations × 4 treatments × 2 replicates). After the conclusion of each behavioral session, all the larvae and lab materials were treated with antifungal disinfectant and all the amphibians were released in their site of origin (see Ethical statement).

Behavioral traits Behavior and activity of larvae were obtained by extracting individual movements from videos with the video-tracking software idTracker. This software allows to track identity and position of multiple animals in subsequent frames of a video, by recognizing individual shape based on their size and chromatic contrast with the background (Pérez-Escudero et al. 2014). We considered three behavioral traits: total distance moved by larvae during the test (hereafter total distance), mean distance from the barrier separating them from the stimulus source (avoidance), and the number of bursts performed by larvae (number of bursts). Two of them, total distance and avoidance, are classical behavioral parameters describing prey activity and space use (Lima and Dill 1990). General decrease of activity and avoidance of risky areas are common antipredator behaviors that are frequently observed in amphibian larvae (Relyea 2001a; Van Buskirk et  al. 2012; Winandy and Denoël 2013; Manenti et al. 2016). However, in preliminary observations, we noticed that some species show periods of limited movement followed by rapid bursts. These bursts lasted few seconds and allowed larvae to cover large distances, a behavior likely representing an escape attempt (Dayton et al. 2005; Teplitsky et al. 2005). Measuring total movement only could have obscured specific antipredator strategies, potentially leading to the misinterpretation of behavioral responses. Thus, for all the species, we considered the number of bursts performed by larvae during each test as an additional behavioral parameter. For each species, we calculated the mean distance moved during single movements (i.e. continuous movements through time,

Melotto et al. • Visual cues and coevolution drive antipredator responses

1355

V – C+

V + C+

V + C–

Visual cues :

V – C–

Chemical cues :

Figure 1 Experimental scheme. The activity of amphibian larvae was tested in behavioral trials with the exposure to four treatments: contemporary presence of visual and chemical stimuli of the American crayfish (V+C+); presence of chemical cues only (V–C+); presence of visual cues only (V+C–); absence of crayfish cues (V–C–). During the tests, tadpoles were housed in one compartment of two-sided experimental tanks (prey compartment). The predator compartment was separated by a transparent plastic barrier and hosted an adult American crayfish. The crayfish was placed in an opaque pot in treatments excluding visual stimuli (V–C+ and V–C– ), while it was free ranging in treatments with exposure to visual stimuli (V+C+ and V+C–). Exposure to chemical cues (treatments V+C+ and V–C+) was allowed by means of small holes in the barrier and in the pot hosting the crayfish; holes were absent in treatments excluding exposure to chemical cues (V+C– and V–C–). Behavioral tests lasted 7 min and each larva (n = 24 individuals per species) was exposed twice to each treatment.

interspersed with periods of inactivity) and its standard deviation (SD). All the movements exceeding the mean movement + 2SD were defined as bursts. This approach allowed detecting rare movements that considerably differed from the average, while ensuring consistency across species. Correlations between the three behavioral traits analyzed showed that avoidance was weakly correlated to total distance or the number of bursts. Total distance was generally positively related to the number of bursts (see Supplementary Table S1). This is not surprising as, generally, larvae covered relatively long distances when performing bursts, thus individuals exhibiting higher bursts frequencies also moved more. However, these behaviors are two distinct aspects of antipredator responses which can represent different antipredator strategies (i.e., avoiding predator detection vs actively escaping once detected), and prey responses can be differentially expressed according to the perceived risk or show different effectiveness depending on predator hunting

strategy (Relyea 2001b; Teplitsky et  al. 2005; Rehage et  al. 2009; Mogali et al. 2011; Ferrari et al. 2015).

Statistical analysis The effects of crayfish exposure on amphibian behavior were analyzed through Bayesian multivariate Generalized Linear Mixed Models (GLMMs). These models allow to consider the influence of fixed effects on the dependent variable while taking into account the covariation between multiple dependent variables and the non-independence of observations (e.g., repeated observations on the same individual or on the same population) (Pinheiro and Bates 2000; Bürkner 2018). We built a multivariate model considering the three behavioral traits (total distance, number of bursts, and avoidance) as dependent variable and simultaneously testing responses of all species. We included treatments (chemical or visual cue exposure) as fixed factors to

Behavioral Ecology

1356

assess the effect of crayfish exposure on larval behavior. Potential non-additive effects between chemical and visual cues were assessed by testing statistical interactions between treatments. Moreover, to test the hypothesis that coevolution with the native crayfish species could increase the ability to recognize the invasive ones, amphibians were classified according to their history of co-existence with the native crayfish (species living in habitats once hosting A.  pallipes vs species exploiting different habitats where the native crayfish never occurred; see Figure 3 and Supplementary Figure S1). History of coexistence with the European crayfish was included in the model as an additional fixed factor, while we used 2-way interactions between coexistence and chemical or visual cues to assess if responsiveness to a particular stimulus from the invasive crayfish was affected by the coevolutionary history with the native crayfish. Air temperature (°C) and daytime (minutes from midnight) were included as covariates, as they can affect amphibian activity (Wells 2007). All continuous independent variables were standardized before analyses. Moreover, a few videos were slightly shorter, thus we also included video duration as an additional covariate in all models. As random factors we included species identity, population of origin, individual identity and test replicate (first or second exposure to a single cue). In mixed models, we took into account the nested structure of random factors (individual, population, and species identity) (Zuur et al. 2009). The multivariate GLMM was run with three MCMC chains using 2000 iterations and a burn-in of 1000 in the brms package in R (Bürkner 2018). For all variables, c-hat was